AI Fintech
AI Procurement Agent
An autonomous agent was developed to manage supplier communications and negotiations directly within the client's CRM.
- Switzerland
- 2025
Outcomes
- Autonomous across WhatsApp and email
- Supplier outreach
- Free text, images, technical documents
- Request format accepted
- Inside amoCRM, every exchange logged
- Where it runs
The Vision & Challenge
A leading Swiss firm in the interior design and renovations industry built its reputation on meticulous project execution. Success in this market depends on coordinating a complex network of material suppliers and specialized craftspeople, from carpenters to metalworkers. Every detail must be managed with precision to meet high client expectations.
The firm's growth, however, was constrained by an internal friction point-procurement. The process of gathering project needs, contacting multiple suppliers for quotes, and negotiating terms was entirely manual. This administrative work was slow, susceptible to errors, and consumed the procurement team's focus, preventing them from concentrating on strategic sourcing and relationship management.
Robust Devs was engaged to design and build a specific tool to solve this operational drag. We engineered an AI-driven agent that integrates directly into their amoCRM platform. This agent interprets unstructured requests from project managers and then autonomously engages with suppliers to secure quotes, effectively closing the loop on a critical and time-intensive business function.
The challenge
- The system needed to interpret complex, unstructured project requirements submitted in various formats, including free-form text, images, and technical documents.
- Automating supplier outreach required managing concurrent, real-time conversations across different channels like WhatsApp and email, while consistently applying negotiation logic.
- To ensure user adoption and workflow continuity, the solution had to function as a native component within the client's existing CRM, not a separate application.
How we solved it
- We built a custom Natural Language Understanding engine to parse mixed-media inputs. The engine extracts key data points-materials, dimensions, deadlines-and structures them for the agent.
- A multi-channel communication module was developed to interact with supplier APIs. It manages conversations as independent threads governed by a configurable rules engine for price and timeline negotiation.
- Using the amoCRM API, we built custom React widgets that embed the agent's interface directly into the CRM. All communications are automatically logged to the corresponding project record for a full audit trail.
Technologies used
Front-end
- JavaScript
- ReactJS
- HTML5
- CSS3
Back-end
- PHP
- Laravel
- Python
DevOps
- Docker
- Github CI
- Digital Ocean
Database
- MySQL
Natural Language Instruction
Project managers can submit procurement requests in their natural workflow, using plain text, attaching plans, or sharing links. The AI engine parses these unstructured inputs, identifies the necessary items and specifications, and converts them into a structured brief. This removes the need for manual data entry and rigid forms, saving time and reducing misinterpretation.
Autonomous Supplier Negotiation
Once a request is processed, the agent identifies and contacts relevant suppliers from the CRM database via WhatsApp or email. It conducts independent conversations to obtain quotes, negotiating on price and delivery dates based on pre-set business rules. All communications are logged and the best quotes are presented, allowing the procurement team to focus on final decisions rather than the negotiations.
Have a project like this in mind?
Tell us what you're building. We'll show you how we'd approach it and what it would take to ship.